[Robotics] Force/torque/collision sensor

Fernando L. Garcia Bermudez fgb at eecs.berkeley.edu
Tue Aug 2 09:30:29 CEST 2011 

Anna Chiara, Arnaud, Benoit,

Sorry for the late response, but research took over for a bit.

Thanks for catching the Matrix.transpose[d]() issue and proposing a fix; I
still don't understand how I overlooked it. I ended up going with
beam.worldOrientation.transposed() for conciseness and have updated the repo
addressing this:

https://github.com/fgb/passive_compliant_joint/commit/05f81a3036a632717e344a7ea0a684f68221f79e

Thanks,

-- Fernando


On Wed, Jun 15, 2011 at 09:40, Benoit Bolsee <benoit.bolsee at online.be>wrote:

> **
> The method that returns a new matrix is Matrix.transposed()
> The API didn't change with regards to that.
>
>  -----Original Message-----
> *From:* robotics-bounces at blender.org [mailto:robotics-bounces at blender.org]
> *On Behalf Of *Anna Chiara Bellini
> *Sent:* mercredi 15 juin 2011 13:21
> *To:* Blender and Robotics
> *Subject:* Re: [Robotics] Force/torque/collision sensor
>
> Hi Fernando,
>
> my issue was this: the mathutils.Matrix*.transpose()*
> (beam.worldorientation.transpose(), line 11 ) doesn't return a new instance
> of Matrix, but rather it trasnposes the matrix itself, and returns None.
> As such, on line 11 the beam orientation is transposed, the operation
> returns None and the multiplication fails with and exception, and the rest
> of the script is ignored
>
> rot = (beam.worldOrientation.transpose() *
> base.worldOrientation).to_euler()
>                                     this ^ fails
>
> On the Blender console I see:
> Python script error - object 'Beam', controller 'Python Script#CONTR#1':
> Traceback (most recent call last):
>   File "joint_controller.py", line 11, in <module>
> TypeError: Matrix multiplication: not supported between 'NoneType' and
> 'mathutils.Matrix' types
>
> So, here is what I did:
>
> beam_rot = beam.worldOrientation.copy() # so that subsequent transpose()
> doesn't change the beam's real orientation
> beam_rot.transpose()
> rot = (beam_rot * base.worldOrientation).to_euler()
>
> I have given it for granted this was a bug introduced by Blender's huge api
> change, but I'm not sure, I wanted to try it on an older Blender, but...
> hem, I don't know how to work with pre-2.5 :-)
>
> Thank you again
>
> Anna Chiara
>
>
>
> On 10 June 2011 19:00, Fernando Garcia Bermudez <fgb at eecs.berkeley.edu>wrote:
>
>> Anna Chiara,
>>
>> Can you provide more info on the issue/change? I'm using 2.57 and not
>> seeing anything wrong, but I might be overlooking it. If you can send the
>> changed script I can test it and update it on github for future users.
>> Thanks,
>>
>> -- Fernando
>>
>>
>>
>> On Fri, Jun 10, 2011 at 02:56, Anna Chiara Bellini <
>> annachiara.bellini at gmail.com> wrote:
>>
>>> Ciao Fernando,
>>> I've been playing around with this for a while, and I think I got the
>>> hang of it, thank you :-)
>>> I just had to change the script a little because the mathutils.Matrix api
>>> has changed in Blender 2.56 (namely: I had to copy() and then .transpose()
>>> the beam rotation Matrix), but it works very well.
>>>
>>> I'm working on the linear springs now, and I'm planning to transform this
>>> in a Morse-style sensor, even though a few peculiarities will apply.
>>>
>>> Thank you again for your kind help
>>>
>>> Anna Chiara Bellini
>>>
>>>
>>> On 7 June 2011 17:40, Fernando Garcia Bermudez <fgb at eecs.berkeley.edu>wrote:
>>>
>>>> Since the e-mail below was being held due to a large attachment, I
>>>> decided to re-post without it. You can grab the attachment from:
>>>>
>>>> http://github.com/fgb/passive_compliant_joint
>>>>
>>>> Thanks,
>>>>
>>>> -- Fernando
>>>>
>>>> On Mon, Jun 6, 2011 at 12:05, Fernando Garcia Bermudez
>>>> <fgb at eecs.berkeley.edu> wrote:
>>>> > Anna,
>>>> >
>>>> > I've constructed passive compliant joints in the past using a mixture
>>>> > of rigid body joints and a python script controller, since there are
>>>> > no generic spring constraints in Blender that I know of.
>>>> >
>>>> > In a nutshell, I create a generic 6 DoF rigid body joint between two
>>>> > bodies, disabling collisions between them. I then proceed to allow
>>>> > only the rotations that the spring joint I'm envisioning has. I setup
>>>> > rigid body physics for both bodies and then add some parameters to the
>>>> > game property editor (spring constant, damping constant, and resting
>>>> > position for the joint). Finally, I add an always sensor connected to
>>>> > a python script which effectively controls the passive compliant
>>>> > joint. The script implements torque feedback based on the game
>>>> > property parameters. You could easily implement force feedback for a
>>>> > linear spring joint in a similar way. For clarity, I'm including a
>>>> > demonstration blend file.
>>>> >
>>>> > As you can see, the forces/torques are available to you as part of the
>>>> > simulation, but as Gilberto notes, they might not be very precise due
>>>> > to Bullet's limitations.
>>>> >
>>>> > I hope this helps you solve your issue. Let me know if you have any
>>>> questions,
>>>> >
>>>> > -- Fernando
>>>> >
>>>> >
>>>> > On Mon, Jun 6, 2011 at 11:09, Anna Chiara Bellini
>>>> > <annachiara.bellini at gmail.com> wrote:
>>>>  >> Hello Gilberto,
>>>> >>
>>>> >> I'll try to explain it better... I need to detect forces for two
>>>> reasons:
>>>> >>
>>>> >> 1 - our robot is an UAV, and as it flies, it will get in touch with
>>>> objects
>>>> >> and environment. On the robot I need to have a few force/torque
>>>> sensors,
>>>> >> that will tell me the force/torque applied in that spot, and then we
>>>> combine
>>>> >> the different values to get a resulting force that will get back to
>>>> the user
>>>> >> using a force-feedback device.
>>>> >>
>>>> >> 2 - the robot features a grasp/manipulate tool, and I need to read
>>>> the force
>>>> >> applied on it.
>>>> >>
>>>> >> Now, my first idea is this: if I model the sensor as a small pivoting
>>>> >> damper, then I could read the position and pose of the damper to
>>>> calculate
>>>> >> force and torque... but this is just a very basic idea, I don't know
>>>> if it's
>>>> >> correct or if there are better ways. (Ideally, this should end up
>>>> being a
>>>> >> MORSE sensor, but that would be a second step, hence why I posted the
>>>> >> question here instead on the morse mailing list)
>>>> >>
>>>> >> I have also found this on the Bullet forum:
>>>> >>
>>>> http://www.bulletphysics.org/Bullet/phpBB3/viewtopic.php?p=&f=9&t=3637,
>>>> but
>>>> >> I'm not understanding it completely, for instance I do not know what
>>>> a
>>>> >> "proxy" object is and how to set it up.
>>>> >>
>>>> >> Hope this is clear enough :-)
>>>> >>
>>>> >> You are saying that the Bullet engine is not extremely precise... can
>>>> you
>>>> >> tell me what are the limitations, or point me to a place where I can
>>>> get
>>>> >> more information?
>>>> >>
>>>> >> Thank you again
>>>> >>
>>>> >> Anna Chiara Bellini
>>>> >>
>>>> >>
>>>> >> On 6 June 2011 18:09, Gilberto Echeverria <
>>>> gilberto.echeverria at laas.fr>
>>>> >> wrote:
>>>> >>>
>>>> >>> Hello Anna Chiara,
>>>> >>>
>>>> >>> For the moment, MORSE does not include this sort of force sensors in
>>>> >>> joints. So far, the main developers of MORSE have not approached
>>>> this sort
>>>> >>> of simulation, so it remains an open area for development. However,
>>>> maybe
>>>> >>> other people in the Blender community may have worked on similar
>>>> problems.
>>>> >>>
>>>> >>> It should be possible to measure forces in Blender, but the physics
>>>> engine
>>>> >>> used by default (Bullet) is not extremely precise. It will depend on
>>>> what
>>>> >>> exactly you want to model.
>>>> >>> If you can provide more details, we can give you a better insight on
>>>> how
>>>> >>> to deal with this kind of sensors.
>>>> >>>
>>>> >>> Best regards,
>>>> >>> Gilberto Echeverria
>>>> >>>
>>>> >>> On 06/06/2011 04:16 PM, Anna Chiara Bellini wrote:
>>>> >>>
>>>> >>> Dear all,
>>>> >>>
>>>> >>> I am working on a robot simulation project (
>>>> http://www.airobots.eu/ )
>>>> >>> with Blender and I am trying to do something which I'm sure someone
>>>> else has
>>>> >>> already done, so I'm asking for directions: I need to model passive
>>>> >>> compliant joints and detect the forces acting on them. I guess that
>>>> the way
>>>> >>> is through Blender's rigid body joint constraint , possibly the
>>>> >>> 6DofSpringConstraint, but I'm not really sure about it and I really
>>>> don't
>>>> >>> know which would be the best solution.
>>>> >>>
>>>> >>> In our project, we are using MORSE for sensors/actuators/middleware,
>>>> so if
>>>> >>> anyone had a MORSE-specific solution or suggestion, that would be
>>>> great.
>>>> >>>
>>>> >>> Any ideas?
>>>> >>>
>>>> >>> Thank you in advance
>>>> >>>
>>>> >>> Anna Chiara Bellini
>>>> >>>
>>>> >>> _______________________________________________
>>>> >>> Robotics mailing list
>>>> >>> Robotics at blender.org
>>>> >>> http://lists.blender.org/mailman/listinfo/robotics
>>>> >>>
>>>> >>>
>>>> >>> --
>>>> >>> Gilberto Echeverria
>>>> >>> Research Engineer
>>>> >>> Robotics and Interactions Group (RIS)
>>>> >>> LAAS/CNRS, Toulouse
>>>> >>> +33 (0)5 61 33 78 95 <%2B33%20%280%295%2061%2033%2078%2095>
>>>> >>>
>>>> >>> _______________________________________________
>>>> >>> Robotics mailing list
>>>> >>> Robotics at blender.org
>>>> >>> http://lists.blender.org/mailman/listinfo/robotics
>>>> >>>
>>>> >>
>>>> >>
>>>> >> _______________________________________________
>>>> >> Robotics mailing list
>>>> >> Robotics at blender.org
>>>> >> http://lists.blender.org/mailman/listinfo/robotics
>>>> >
>>>> _______________________________________________
>>>> Robotics mailing list
>>>> Robotics at blender.org
>>>> http://lists.blender.org/mailman/listinfo/robotics
>>>>
>>>
>>>
>>> _______________________________________________
>>> Robotics mailing list
>>> Robotics at blender.org
>>> http://lists.blender.org/mailman/listinfo/robotics
>>>
>>>
>>
>> _______________________________________________
>> Robotics mailing list
>> Robotics at blender.org
>> http://lists.blender.org/mailman/listinfo/robotics
>>
>>
>
> _______________________________________________
> Robotics mailing list
> Robotics at blender.org
> http://lists.blender.org/mailman/listinfo/robotics
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/robotics/attachments/20110802/95354184/attachment.htm

More information about the Robotics mailing list